Discussion Optane Client product current and future

[nosirrahx]

Yes, that looks like it would be RAM cache (though the 4K QD1 Read is low for RAM).

I think Intel is aware that since 4K is exactly what it says, super small files, there really is no need for crazy 4K speed as Optane already delivers 4K data far faster than any NAND solution.

Large files by contrast benefit from the bandwidth. To me it looks like Intel is devoting some kind of RAM caching only for the files that would benefit from it the most and leaving the smaller files to Optane alone.

In any event this laptop is crazy fast no matter what is happening under the hood. Boots are blazing and apps load instantaneously. If Optane is doing this it is easier to make a case for 16GB or RAM over 8GB as I am sure that the cache is based on free RAM.

 

[IntelUser2000]

I think Intel is aware that since 4K is exactly what it says, super small files, there really is no need for crazy 4K speed as Optane already delivers 4K data far faster than any NAND solution.

This is my guess too. RAMdisks are not that much faster at random read 4K compared to NVMe Optane. There's really no point in faster random reads for storage systems. It's even questionable whether Optane's perceived responsiveness improvement over NAND SSD is due to ultra low latency(and high random read) or that it doesn't need a DRAM cache, and doesn't suffer from drive being full or dirty.

The first thought I had in mind was "Primo Cache patch!"

I have been checking out laptops with 22110 support, not much to choose from. Once the M.2 905P drops I might see about creating a 860 EVO + 905P laptop to get the ultimate in performance and space.

That's because you sir, are a rare breed.

 

[cbn]

Large files by contrast benefit from the bandwidth. To me it looks like Intel is devoting some kind of RAM caching only for the files that would benefit from it the most and leaving the smaller files to Optane alone.

Yeah, that makes sense......with the "larger files" I suspect being mirrored from the Optane cache to the RAM cache.

 

[cbn]

Tom's hardware predicts Optane NVMe + 3D QLC NVME:

https://www.tomshardware.com/news/intel-optane-memory-qlc-cache,37223.html





My biggest concern with 3D QLC (like it was with planar TLC) is data retention.

P.S. I am assuming the author is thinking about 16GB Optane paired with 512GB 3D QLC NVMe at the lowend (but I believe the 16GB Optane 4K QD1 write will be much slower* than 512GB 3D QLC NVMe. Sequential write will definitely be much slower).

*The same amount of Optane (16GB) as DDR interface would change this relationship to 16GB Optane 4K QD1 write >> 512GB 3D QLC NVMe 4K QD1 write. (I am concerned though how long till we see such a low end configuration). Could it be that one way to speed adoption of low end DDR interface Optane is to push more than four bits per cell into 3D NAND. This or more aggressive than usual lithography shrinks + (possibly) less planes per die? (And this is one reason why Intel is diverging from Micron for 3D NAND development?)

With respect to what I bolded above it looks like the relationship would be 32GB NVMe Optane* + 512GB 660p vs. 512 GB760p.

Reason: The 512GB 660p (3D QLC) drive is only $99.99 to the 512GB 760p's (3D TLC) $159.99. (This $60 price gap allowing, of course, for the inclusion of the $59.22 32GB Optane)

*not 16GB Optane

 

[IntelUser2000]

Data retention shouldn't be an issue.

The JEDEC spec for 1-year retention on client and 3-months retention for servers only apply to SSDs after their write endurance has been exhausted.

Generally its believed that it'll retain its data for 10 years.

 

[IntelUser2000]

More data about P4801X, so it would be an interest to potential 905P M.2 buyers.

P4801X is called "Low Power Optane". To be available in 100GB/200GB/375GB sizes. P4801X is also coming in the U.2 form factor, but only at 100GB. Performance for writes vary significantly based on capacity. Power will be somewhat lower.

It seems it may have been delayed.

 

[cbn]

Linus Tech Tips really likes Optane with Primocache :

https://youtu.be/rWXBo0bb_dU?t=445

 

[IntelUser2000]

Laptop review with Optane Memory M10: https://www.notebookcheck.net/HP-14...Memory-HD-Display-Laptop-Review.328732.0.html

Seems pretty efficient. The battery life is actually pretty decent considering.

 

[cbn]

Laptop review with Optane Memory M10: https://forums.anandtech.com/threads/optane-client-product-current-and-future.2525180/page-11

Seems pretty efficient. The battery life is actually pretty decent considering.

IntelUser2000, that link doesn't lead to a laptop review......but I am guessing you must be referring to this review (Acer Aspire 5).

Battery life
We measure battery life in a laptop by looping a 4K video using the stock Window 10 video player. We set screen brightness as close to 250 nits as we can (which meant cranking the brightness all the way on the Aspire 5’s dim display), with the volume set at 50 percent and headphones plugged in.

The Acer Aspire 5’s battery drain time of 519 minutes (a little over 8.5 hours) is close to what we’ve seen from other laptops with similar 48 watt-hour batteries, so no surprises here. Yes, the Dell New XPS 13 manages to squeeze a lot more juice out of a slightly larger (52 watt-hour) battery, but the Dell costs more than twice as much as the Aspire 5.

Yep, that does look pretty good considering the Acer E15 (E5-576-392H) uses 1TB HDD and the Acer E15 (E5-576G-5762) uses 256GB M.2 SATA SSD.

With that noted, I wonder how the laptop would do with light browsing.

 

[cbn]

Regarding the laptop test in the previous post.....with video playback the system would be bypassing the optane (based on what is written below).

http://surprizingfacts.com/what-is-intel-optane-part-1-optane-memory-blog-of-intel-surprizingfacts/

Secondly, while the RST system is running, the driver will continuously generate caching. And here there is one important difference between Optane Memory modules of different capacities: on a 16GB device, only block level caching is supported, on a 32GB device, block level caching and file leveling (both work simultaneously). In the case of block caching, the decision to cache a block occurs instantly at the time of the I / O request. In the case of file caching, the driver monitors the frequency of access to files and puts it all in a special table, which then (at the time of system downtime or according to the user's schedule) is used to determine which files are left in the cache, which are deleted and which are added.

Both types of caching use rather clever algorithms for caching decision making – I can not describe them deeply, but for general understanding I note that, for example, video files are not cached (yes, the driver looks at File extension), the size of the file is taken into account, the type of load is determined – caching preferences are given to random access rather than sequential, which makes sense due to extremely slow operation of hard disks in random access operations, etc. On the Internet, I met some negative comments on the topic that "the cache will be immediately overwritten with data," "16GB capacity is not enough for anything" and the like – usually reviews from people who have never tested Optane Memory. I have not yet heard negative feedback about the performance of this solution from any of our partners I work with.

....So power consumption based solely on HDD power consumption, not Optane (active power).

 

[IntelUser2000]

Ah, thanks. I edited the post. It was from Notebookcheck. But nice one from PCWorld too.

The competing Acer laptops use a 6-cell, 60WHr battery. 33% larger in capacity than the Aspire 5.

You have to note that its using a 16GB version. It may be M10, but the algorithm may be based on the 16GB. Also if its allowing the hard drive to power down, it would explain the advantage. To be comparable to an SSD system, not only Optane module has to use full breadth of its power save capabilities, but also allow the HDD to power down. A feat that's often hard to see.

The E5-576G-5762 is not directly comparable as its disadvantaged by the discrete GPU.

So these are examples where testing instantaneous power use or sticking it in an older H-class CPU doesn't work. The NVMe power down features must be working otherwise the M10 module itself would use 1W doing nothing. Some documents also say the M10 only has full features when using on the 8th Gen core platforms.

 

[nosirrahx]

Are there any laptops that come with the 32GB module?

 

[IntelUser2000]

Can't find a laptop with a 32GB module. I think its because SSDs are more natural in laptops so there's little point shelling extra for the 32GB version.

I've seen some desktops with 32GB though.

 

[nosirrahx]

Can't find a laptop with a 32GB module. I think its because SSDs are more natural in laptops so there's little point shelling extra for the 32GB version.

I've seen some desktops with 32GB though.

For lower end laptops I guess but they do sell laptops with 4TB of storage. A 4TB SATA SSD + 32GB Optae is cheaper than 2 2TB NVMe SSDs.

2 2TB 970 EVOs are $1350. A 4TB 860 EVO is $900 and with a 32GB Optane cache drive it will usually even feel faster.

 

[cbn]

You have to note that its using a 16GB version. It may be M10, but the algorithm may be based on the 16GB. Also if its allowing the hard drive to power down, it would explain the advantage. To be comparable to an SSD system, not only Optane module has to use full breadth of its power save capabilities, but also allow the HDD to power down. A feat that's often hard to see.

I got the impression looking at the explanation below that even with the 16GB using block level cache (not file block level + file level cache as the 32GB does) that it still looks at the file extension.

http://surprizingfacts.com/what-is-intel-optane-part-1-optane-memory-blog-of-intel-surprizingfacts/

Secondly, while the RST system is running, the driver will continuously generate caching. And here there is one important difference between Optane Memory modules of different capacities: on a 16GB device, only block level caching is supported, on a 32GB device, block level caching and file leveling (both work simultaneously). In the case of block caching, the decision to cache a block occurs instantly at the time of the I / O request. In the case of file caching, the driver monitors the frequency of access to files and puts it all in a special table, which then (at the time of system downtime or according to the user's schedule) is used to determine which files are left in the cache, which are deleted and which are added.

Both types of caching use rather clever algorithms for caching decision making – I can not describe them deeply, but for general understanding I note that, for example, video files are not cached (yes, the driver looks at File extension), the size of the file is taken into account, the type of load is determined – caching preferences are given to random access rather than sequential, which makes sense due to extremely slow operation of hard disks in random access operations, etc. On the Internet, I met some negative comments on the topic that "the cache will be immediately overwritten with data," "16GB capacity is not enough for anything" and the like – usually reviews from people who have never tested Optane Memory. I have not yet heard negative feedback about the performance of this solution from any of our partners I work with.

 

[IntelUser2000]

Actually, you are right. Thanks for the correction.

I wonder what happened to the 905 M.2?

 

[nosirrahx]

I wonder what happened to the 905 M.2?

I have been trying to find confirmation on release dates for months, there does not seem to be one.

I have a suspicion that it might have been canceled. I want to buy one to experiment with.

I want to benchmark my travel laptop with the 800P VS the 905P in an Optane cache scenario.

I will try asking on the Intel support forum, but I think we all know how that will go.

https://communities.intel.com/message/566798#566798

 

[cbn]

While looking for 905p M.2 I found out that the 280GB 900p Add-in card is $260 after promo code:

https://www.newegg.com/Product/Product.aspx?Item=N82E16820167437&ignorebbr=1

Wow, that is a really good price!

And 32GB Optane is $42.99 plus $1.99 shipping:

https://www.newegg.com/Product/Product.aspx?Item=N82E16820167427

Another really good price!

 

[nosirrahx]

While looking for 905p M.2 I found out that the 280GB 905p Add-in card is $260 after promo code:

https://www.newegg.com/Product/Product.aspx?Item=N82E16820167437&ignorebbr=1

Wow, that is a pretty good price!

Very good and could be used as either an OS+apps drive or as a system accelerator for a huge primary drive.

2TB 860 EVO is $400 + $260 for 900P is slightly cheaper than a 2TB 970 EVO but way faster.

 

[cbn]

Very good and could be used as either an OS+apps drive or as a system accelerator for a huge primary drive.

2TB 860 EVO is $400 + $260 for 900P is slightly cheaper than a 2TB 970 EVO but way faster.

That is a really good comparison.

And with the 512GB Intel 660p at $99.99 (launch price at Newegg) I have to imagine that a 2TB version of Intel 660p would also be around $400 (if not less).

So I have to wonder how long till we compare Optane + NVMe 3D QLC vs. NVMe 3D TLC for the larger capacities.

P.S. Here was my comparison of 32GB Optane + 512GB 660p (NVMe 3D QLC) vs. 512GB 760p (NVMe 3D TLC)

 

[cbn]

Maybe even better than Optane + NVMe 3D QLC (with dram buffer) would be Optane + NVMe 3D QLC (dram-less, but using host memory buffer).

Then if say the system begins paging out and thus not having RAM available for host memory buffer the performance impact would be zero to minimal. (perhaps only wear on the 3D QLC NAND being the difference).

 

[Dayman1225]

Update on the M.2 905P, it will be arriving in October

 

[cbn]

The E5-576G-5762 is not directly comparable as its disadvantaged by the discrete GPU.

The discrete GPU doesn't appear to be hurting it when compared to E5-576-392H.

So I am thinking the disadvantage is probably coming from the E5-576-392H having the 1TB 2.5" HDD vs. the .E5-576G-5762 SATA SSD.

Reason: SATA SSD has better performance per watt compared to 1TB 2.5" HDD.

(So now that we know 16GB Optane (on other laptops) uses HDD for video playback I wonder how much improvement in performance per watt for 2.5" HDD would be needed to compete against SATA SSD?)

 

[IntelUser2000]

Update on the M.2 905P, it will be arriving in October

Nice find. It also seems they added a heatspreader so it looks like the regular 905P drive controllers.

Sequential performance drop will be noticeable on the lower capacities, none on the largest. I've detailed the reasons in my other thread. Essentially Optane is power limited.

Edit:

(So now that we know 16GB Optane (on other laptops) uses HDD for video playback I wonder how much improvement in performance per watt for 2.5" HDD would be needed to compete against SATA SSD?)

Don't forget the other Acer systems have a 33% larger battery compared to the Optane Memory-using Aspire 5. So even for video playback Optane Memory is quite efficient. So I do not believe its simple as "its running on the HDD with video playback".

You have to remember even against the 2.5" HDD, it needs to overcome significant hurdles just to have equal battery life. It's HDD + Optane versus HDD. In order for the former to be noticeably lower power than the latter it requires Optane doing most of the work, going quickly to idle when not needed, and idle power management working. That's a very difficult trifecta of requirements that often isn't met with hybrid drive systems.

Not to mention, platter HDDs are actually quite slow in ramping down. It needs to be idle for a significant period(seconds to mins) before HDD idles.

 

[Dayman1225]

Looks like Intel plans to up the capacity of the U.2 Version of the 905P to 960GB and 1.5TB and upto 765GB for the M.2 version.

Source